Which samplerate for your DAW? [a sort of poll]

voidar · Post by **voidar** » Thu Oct 19, 2006 5:20 am

The benefits of 32-bit increases with higher sample-rates. More sample-points.

I use 44.1KHz 24/32-bit these days. I would love to use higher rates, but with 12DSPs I need to stick to 44.1KHz.

darkrezin · Post by **darkrezin** » Thu Oct 19, 2006 9:02 am

44.1 khz, 24 bit here.

Advantages of higher rates are not enough to justify other disadvantages (for me anyway).

erminardi · Post by **erminardi** » Thu Oct 19, 2006 9:28 am

Ok, there are 2 different point here:

1) 44.1 24 bit seems to be the best solution for recording DAW

2) 96 is the ideal for the sound generation (better filter sweep, more presence & clarity, etc.)

I have 2 different PCs, one as DAW and one as external syshtesizer module. The only problem is that with ADAT connection I must set the same resolution for both PCs.
If I want the Synths PC at 96khz I must transfer the audio not via ADAT but via analog cables, this means more A/D & D/A converter, a lot of more cables (& noises) and a less versatile system communication.
Maybe this trick don't worth the little audio benefit...

Anyway the best solution could be to (re)make new synthesizers with internal oversampling like creamware ones.

gainman · Post by **gainman** » Thu Oct 19, 2006 10:03 am

I'm using 48 KHz, without any good reason for that. I might move to 44.1 KHz...

erminardi · Post by **erminardi** » Thu Oct 19, 2006 10:37 am

Obviously the oversampling trick is not the holy graal for a well sounding synthesizer... Uknow, Miniscope and many other standard little flat devices sounds like shit also @ 96khz!!!

From free synth arsenal I like Orbitone three-o-three, Saturn [edit], Amphetamine, Adern Kick me, and Wavelength/Hummel's stuff.
The rest is almost useless...
Look also for some very interesting modular patches like i.e. Morpheus and Morpheus2.

A lot of people in KVR hates Scope and says that it sounds like native VST because they have hear only the standard synthesizer bundled with LunaII!!!
A friend of mine gave to me his PulsarII for a piece of cake for this reason...

[ This Message was edited by: erminardi on 2006-10-20 01:31 ]

wayne · Post by **wayne** » Thu Oct 19, 2006 5:10 pm

erminardi - a bit nitpicky perhaps, but the Saturn is just "Saturn" by Rainer Hain - from 2002. Jngaelin is a Z member who loves the synth and made some presets.

[pedantic mode off]

Liquid Len · Post by **Liquid Len** » Fri Oct 20, 2006 3:50 am

Generally I use 48khz and 32 bit recording. I would probably use 24 bit but I read somewhere that Cubase, or ASIO, reads 2 bytes at a time so 24 bits takes just about as much processing power as 32 bits. Dunno if that is true (can anyone confirm this), 5 years later, but since then I've been using 32 bits mostly, hard disk space is not an issue. I don't mind using 48k instead of 44k, 96k is very obviously higher quality but it takes too much processing power. (And I don't have a 96khz AD convertor except for the onboard analog ins).

I think it's misleading to point out that audio cds are 16 bits, its like doing calcs where you hang on to the extra precision and round off only at the end, the extra precision will give you a more accurate result. Though probably more important to accurately preserving the overall sound is to use a creamware mixer and not the cubase mixer - some signal summing is more equal than others.

astroman · Post by **astroman** » Fri Oct 20, 2006 5:56 am

On 2006-10-20 04:50, Liquid Len wrote:
...I think it's misleading to point out that audio cds are 16 bits, its like doing calcs where you hang on to the extra precision and round off only at the end, the extra precision will give you a more accurate result. ...

well, imho it would be easier to distinguish between sound and precision
they don't correlate necessarily, at least not in practical terms...

a higher numeric spec of a converter tells about 'steps', but nothing at all about how precisely those can be reproduced.
even worse, it's a 2 dimensional problem, as the (correct) point in time matters as well as the (precise) amplitude of the signal - regardless if sampled or reproduced.

imho (to make it short and a bit provocative) not a single 'affordable' 24 bit converter can reproducably deal with it's resolution, nor can any (selfcontained) card have a clock beyond 48khz which is stable enough for a sample resolution > 16 bit.
- unless it's supported by special circuitry (onboard or via an external generator).

So to say (just an example) an EMU 192khz 24bit card CANNOT produce precise results according to the physics it's based on.
If you don't believe it, simply calculate the required clock stability and the voltage difference between 2 bits...

from this point of view such a card produces an almost random signal - and it's in no way precise.
The why doesn't it sound like complete crap ?
(I never heard one - I just assume they sound 'decent')

at least with the time based errors, the high sample rate pays off: aliasing artifacts are way off beyond the audible spectrum and don't mess the 'regular' signal.
regarding amplitude errors they may statistically balance each other out, similiar to a dither process.

bottomline: such a system might sound really great, but despite it's impressive figures, it's in NO WAY precise

so why bother about numbers ?

on the other hand a well designed 'outdated' (from specs) system might perform much more precisely as physical tolerances are lower and hence yield a great sound, too.

The legendary Philips TDA-1541 CD converters, blackface Adat (according to hearsay) and A16 fall in this range (imho).

cheers, Tom

[ This Message was edited by: astroman on 2006-10-20 08:11 ]

Gordon Gekko · Post by **Gordon Gekko** » Fri Oct 20, 2006 6:14 am

anyone working below 192 khz/64 bit is a pussy

Liquid Len · Post by **Liquid Len** » Fri Oct 20, 2006 6:43 am

I can see that issues discussing QUALITY of sound are a very touchy thing on this board, as usual. I don't care what ANYONE says - I can hear a difference between lightwave, say, running at 44khz and at 96khz. I just try to strike a balance between quality of sound and practicality.

erminardi · Post by **erminardi** » Fri Oct 20, 2006 7:04 am

On 2006-10-20 07:43, Liquid Len wrote:
I can see that issues discussing QUALITY of sound are a very touchy thing on this board, as usual. I don't care what ANYONE says - I can hear a difference between lightwave, say, running at 44khz and at 96khz. I just try to strike a balance between quality of sound and practicality.

Agree, maybe the recording (sampling) quality, for average ears, is not so differrent at 44.1 48 & 96.
All that I mean is that the "action" of effects and synthesizers (native or DSP) sounds a lot better at higher samplerate.
i.e. 48 sounds unequivocally better than 44.1 so 96khz...

The actual problem is the compromise between the system resources usage and the realism of the sound (I continuously compare Scope synths with my Juno60, Crumar DS2, etc.)

In the native VST systems the situation of compromise is the same but more heavy and variable (every day = an upgrade!!!) to spare CPU power, DSP users are more lucky, in this sense: the Scope quality is at this high level since end of '90!!!

Maybe in future when 192 khz/64 bit becomes an easy standard, with GigaShark DSP, 30Ghz CPU, 100GB of ram and terabytes of HD, we could laugh about the actual "poor" chance of fidelity

[ This Message was edited by: erminardi on 2006-10-20 08:11 ]

astroman · Post by **astroman** » Fri Oct 20, 2006 7:30 am

On 2006-10-20 07:43, Liquid Len wrote:
I can see that issues discussing QUALITY of sound are a very touchy thing on this board, as usual. ...

I don't find it touchy at all

My post was in no way targeted at your listening - I just used your 'precision' comment as a startpoint.

Imho numeric figures and specs (on their own, out of context) are usually overestimated.
Some gear sounds good, some bad, some mediocre.

Yet it is a matter of fact that 96k 'sounds' better because aliasing artifacts cannot interact with the original signal and not because there are more datapoints

cheers, Tom

Liquid Len · Post by **Liquid Len** » Fri Oct 20, 2006 7:40 am

I wasn't referring to your more useful and to-the-point post. 96Khz is not a useful thing for many/most in the creamware world but a discussion of where quality gets lost (when the signals are summed with a mixer? when effects are applied? at point of AD conversion?) would be educational, at least. We all make the call on what is a waste of time, I'm not interested in making that call for someone else. Really, a lot of arguments on this board, when taken to their logical conclusion, indicates we're all a bunch of spoiled idiots, using something more expensive than a soundblaster live card, when technically that's all you need to make good music.

I probably have a different outlook than most here, I'm not a professional musician, I have a day job, this is just a hobby that's grown out of all proportion

So instead of considering what is marketable, I just make music that I enjoy to hear myself (theoretically

)

On 2006-10-20 08:30, astroman wrote:
I don't find it touchy at all
My post was in no way targeted at your listening - I just used your 'precision' comment as a startpoint.
cheers, Tom

[ This Message was edited by: Liquid Len on 2006-10-20 08:51 ]

symbiote · Post by **symbiote** » Fri Oct 20, 2006 11:44 am

sampling rate influences the frequency range that can be encoded and reconstructed.

if the material you record doesn't contain anything above 22050hz, using a sampling rate above 44.1khz isn't very useful. on the same converter, with such material, without processing, 44.1khz and 96khz will sound exactly the same.

processing will benefit from sampling only if the processing you use changes the frequencies that are present in the recording/signal. generally these are modulation effects, like phasing, flangers, pitch modulations (or playing samples at a higher pitch than recorded at), ring-mod, etc. these effects might change the frequency-content of the signal and might push some into aliasing range.

stuff like eq and compressors might not benefit as much, since they only change amplitude and phase of the signal (the proper-behaved ones at least), and don't introduce new frequencies. stuff that models analog gear might be as well-behaved and might get more of a benefit from the higher sampling rate.

also tom is 100% right when he says that specs and numerical stuff is almost meaningless, since it is. something with high numerical specs might not sound very good, while something with lower specs but very good circuitry has a much higher chance of sounding good. a lot of the 64/192 stuff is pure marketing-kwak since to get proper precision at 192khz, you need a clock circuit that costs at least a nut, sometimes two.

personally i use 44.1/24 or /32 depending on which app i'm running.

astroman · Post by **astroman** » Fri Oct 20, 2006 12:53 pm

On 2006-10-20 08:40, Liquid Len wrote:
...I'm not a professional musician, I have a day job, this is just a hobby that's grown out of all proportion So instead of considering what is marketable, I just make music that I enjoy to hear myself (theoretically )...

we have something in common there, but your tracks are much better than mine...
but then I outperform classic composer Schubert easily - he has only one unfinished piece of art

adding to symbiote:
quality gear can become quite expensive if one is after top specs.
I have an old A16, which features 48 opamps of the 5532 type. This is an analog part of outstanding performance (the 5532), if one considers it's single quantity consumer price of 1.5 Euro. For this reason you'll find it almost everywhere, even in 5 figure Euro consoles.

That's exactly why audiophiles usually want to replace it immediately (when modding gear) - it's simply not exclusive enough

A replacement with a reasonably different part (to justify the effort) is about 10 times more expensive.

A (to be) modded A16 would thus require a humble 720 Euro for new opamps alone.
As it wouldn't make much sense to couple the analog stages with crappy electrolytic caps, you'd also want 96 MKP foil types (2 per stage) in the 5 to 10 mF range, adding up to about 300 Euro (3 Euro/piece).
The 348 + 192 positions to desolder and of course to re-solder with new parts in place would of course be filed under fun

but there's one big caveat: the circuit design doesn't match huge foil caps...
Of course manufacturers get better prices then a consumer for their bigger orders, but still the increase of production costs would be astronomical.

imho such a no-compromise design plus an ultrastable clock is the minimum requirement for higher sample rates and resolution - at least as long as vocals and physical instruments are included (in one or the other form).
On a 'pure' electronic production it's of course much easier to benefit from more headroom and higher resolution - but then who knows (or can tell) how precise the math models are ?
Only listening can tell...

btw my Midiman Audiobuddy elcheapo preamp just refused to pass bass guitar properly.
It didn't even care about it's specs which were supposed to easily deal with that frequency range - just a somewhat blunt sound similiar to old strings...

After changing the electrolytic caps against foil types the sound became indeed much more defined, but that was a pita on that tiny (and undocumented) circuit board. On the other hand I must admit that I seem to enjoy the recreational effect of those excursions with the soldering iron...

cheers, Tom

[ This Message was edited by: astroman on 2006-10-20 14:00 ]

Lima · Post by **Lima** » Fri Oct 20, 2006 10:52 pm

So to say (just an example) an EMU 192khz 24bit card CANNOT produce precise results according to the physics it's based on.
If you don't believe it, simply calculate the required clock stability and the voltage difference between 2 bits...

Astro, is this also true for our board? If the answer is "yes", we teorethically don't have any advantage using 24 bit resolution, right?

symbiote · Post by **symbiote** » Fri Oct 20, 2006 11:58 pm

24bit definitely sounds better with pulsar.

husker · Post by **husker** » Sat Oct 21, 2006 1:43 am

Never mind the bits or sample rate, X-Fi is the answer!

p.s. 44.1/24 here...

Mr Arkadin · Post by **Mr Arkadin** » Sat Oct 21, 2006 2:04 am

i love that graph. It proves that X-Fi is the way to go - you can't go against scientific proof like that. Even mp3s can go beyond your current studio quality. Brilliant.

astroman · Post by **astroman** » Sat Oct 21, 2006 2:50 am

Lima wrote:
So to say (just an example) an EMU 192khz 24bit card CANNOT produce precise results according to the physics it's based on.
If you don't believe it, simply calculate the required clock stability and the voltage difference between 2 bits...
Astro, is this also true for our board? If the answer is "yes", we teorethically don't have any advantage using 24 bit resolution, right?

of course it is (basically) true for ANY board, I just picked the EMU for it's extreme position in budget versus pretended (advertized) capabilities.

but your conclusion puts the (original) statement out of context
Given there would be a perfect signal generator, then I am convinced that not a single 24 bit converter exists, which will be able to produce 2 identical sampling runs in the lower 4 bits because the signal interval is simply too low to be physically measured.
Ideally the samples would 'oscillate' around the 'real physical' values, making it a kind of dithering, which may even 'sound' very good.

but it is in no way precise

recently numeric precision was frequently associated with reproduction quality and 'good sound' - I don't think it's that simple.
But in any case I wanted to get away from specs focussing instead of listening.

'Our' cards sound good because they are designed in a certain way, driven by a specific software - and no question that a 24 bit headroom is simply a matter of convenience

cheers, Tom